Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces Grevo, a unified generative recommendation framework that innovatively treats semantic identifiers (SIDs) as evolvable discrete variables, adapting them based on user behavior rather than relying on a static tokenizer. By integrating behavioral SID generation and semantic SID grounding into a single multitask recommender, Grevo eliminates the need for a second learnable model and complex optimization schedules, resulting in a more stable and efficient indexing process. Experimental results across various real-world datasets show that Grevo significantly outperforms existing state-of-the-art generative recommendation methods, highlighting its effectiveness in bridging the gap between content reconstruction and user prediction.
Grevo redefines generative recommendation by evolving semantic identifiers through user feedback, eliminating the need for complex model coupling and optimization.
Generative recommendation has recently emerged as a promising paradigm that reformulates retrieval as autoregressive generation over semantic identifiers (SIDs), achieving strong performance and drawing increasing attention as an alternative to matching. Despite this progress, SIDs are typically frozen by a content-based tokenizer before the recommender is trained, leaving a persistent gap between what best reconstructs an item's content and what a recommender can predict from user behavior. Recent end-to-end methods close this gap by jointly training the tokenizer and the recommender, but coupling the two destabilizes the identifier space and requires a second learnable model, alignment losses, and usually a delicate alternating-optimization schedule. To address this issue, we propose Grevo, a unified Generative recommendation framework with evolutionary item indexing, which treats the SID assignment itself as an evolvable discrete variable that adapts to behavioral feedback rather than as a tokenizer to be trained. Grevo builds on a single multitask recommender that unifies a behavioral SID generation task and a semantic SID grounding task, letting the recommender absorb the tokenizer's role. Through evolutionary item indexing, Grevo then uses the trained recommender itself as a posterior evaluator to reassign a budgeted set of high-risk identifiers under a fixed vocabulary and length. Together, these components turn index construction into a stable, feedback-driven search that requires no second learnable model, no alignment losses, and no alternating-optimization schedule. Extensive experiments on multiple real-world datasets demonstrate that Grevo consistently outperforms state-of-the-art generative recommendation methods.